If you are afraid of "snooping" then why would you use signatures ? A "snooper" is a passive eavesdropper, who wants to see the data but certainly not to make you aware that your emails are inspected. They won't alter emails, or send fake emails, which are the kind of things that signatures can help against.
If you want to defeat such sniffing, then you need to use end-to-end encryption. This will indirectly involved signatures, but on the certificates, not on the emails themselves. To prevent a man in the middle attack, you want to make sure that when you encrypt an email, you are using the true public key of the intended recipient -- that is where certificates and, crucially, signatures from CA, enter the mix. But these signatures predate the actual email; none of them makes any guarantee as to the email contents.
Now if you want to defeat active attackers, then indeed signatures become relevant. You will want to sign your emails so that the recipient may reliably detect whether your emails were altered in transit (this works, of course, only if the recipient expects your emails to be signed, and will suspect foul play if he receives an unsigned email allegedly sent by you).
At the legal level (though it depends a lot on the jurisdiction), all the "non-repudiation" thing is about shifting the burden of proof. An email, like any piece of data, can be used as proof during litigation; whether it is acceptable in that role depends on the context, and arguments which are then produced by both claimants. Basically, one party (let's call it A) will claim that the email really originates from the other party (B) and thus its contents "bind" B in some way (e.g. contractually); the other party (B) claims that the email is fake and was forged (by A or some other miscreant). Arguments include analysis of headers and log files on all involved machines (B's desktop system, email servers...) and possibly other elements (such as eye witnesses who saw B type the email, or other things like that).
What the legal framework that some countries are trying to build are ways to define the default interpretation: namely, is the email reputed genuine by default (and then B has to find good arguments showing the forgery), or is the email reputed unreliable by default (in which case A must do all the proof work) ? Digital signatures, in a cryptographic sense, can be a part of a system (both technical and legal) which, in some countries, will put the burden of proof on the forgery claimant ("B" in my example).
(As an illustration, in Hollywood, a handshake is considered a binding "signature" for a contract, without involving any computer or even any piece of paper with some ink on it. Eye witnesses are used if a dispute arises.)
Thus, signing emails incurs a "legal risk" only if you intend to renege on the email contents, and that legal risk is increased (with regards to an unsigned email) only within a context where you feel confident that you could successfully deny having sent an unsigned email. In practice, emails leave traces in many places when they hop from server to server, so reneging on some email is already "risky", regardless of whether the email was signed or not.
SSL/TLS, in the context of emails, is used mostly opportunistically. You may have it when sending the email to your outgoing SMTP server; recipient may use it when reading his mailbox (e.g. IMAPS, or some SSL-powered Webmail); SMTP servers may try to use SSL when talking to another SMTP server. Though SSL includes signatures, these are computed over technical "challenges" related to the SSL mechanisms, and do not cover the exchanged data. In that sense, SSL-level signatures cannot "technically" be used as proof about the data contents (this is authentication, not signature).
However, legal matters do not reduce to mere technical elements. The context rules. Some may argue, for instance, that generalized use of SSL for server-to-server connections makes it more difficult for outsiders to actually intercept the emails and alter them. So one may argue (specifically, may argue in court) that SSL tends to make forgeries harder. In any case, all this SSL happens for data transit only; intervening servers still see the email "as is" and maintain their log files, which appear to be sufficient to convince judges in practice.
The best defence strategy of a claimant who wants to renege on his emails might be to assert that his own computer is under control of some malware, which sent the email without his consent or even awareness. One may note that this defence "works" regardless of whether the email is signed or not, since the malware can use your private key to sign such forged emails.
To sum up: only an analysis of the complete situation can tell you whether your "legal risk" is impacted, in which direction and to what extent, by your signing your outgoing emails with S/MIME. However, one may surmise that, usually, S/MIME signatures don't change things much for the sender. For the recipient, signatures may lower the risk substantially in the following sense: if the recipient expects millions of incoming emails (e.g. the emails are actually registration contracts for some service), then he will realistically expect hundreds or even thousands of attempts at reneging emails. Generalized use of digital signatures may be expected to streamline the litigation process, thus lowering the involved costs.
If your enemy is the NSA, then all these legal talks are completely irrelevant: legal proofs are good only when you can show them to a judge. When you wake up, tied on an interrogation chair in the basement of some secret facility in South Dakota, there is no judge to talk to...
